I use the code with a small modification, sending already cropped face images to it and skipping the face detection using dlib. This works well on a single face. However, in images with multiple faces or in a video stream, the algorithm never works beyond the first instance. It seems the pupils are not being found after the first.
Is this a recognizable problem? I suspect it has something to do with thresholding but I do not know how to go about and resetting the thresholds
I use the code with a small modification, sending already cropped face images to it and skipping the face detection using dlib. This works well on a single face. However, in images with multiple faces or in a video stream, the algorithm never works beyond the first instance. It seems the pupils are not being found after the first.
Is this a recognizable problem? I suspect it has something to do with thresholding but I do not know how to go about and resetting the thresholds